the role of interest rates and productivity shocks in ... · since in this exercise interest rate...
TRANSCRIPT
The Role of Interest Rates and Productivity Shocks in
Emerging Market Fluctuations
∗
Mark Aguiar
University of Rochester
Gita Gopinath
Harvard University and NBER
April 6, 2007
Abstract
In this paper we use a quantitative model to explore the potential frictions that
distinguish emerging market business cycles from developed small open economies. Fol-
lowing Aguiar and Gopinath (2007) we allow total factor productivity (TFP) to have a
stationary and an integrated component. We also allow for shocks to the consumption
and investment Euler Equations that operate through the interest rate. These “wedges”
represent changes in the intertemporal marginal rate of transformation, which may be
due to changes in observed interest rates, unobserved borrowing constraints, or other
financial frictions. We estimate the model using data from Mexico and Canada. We
show that interest rate shocks orthogonal to domestic TFP fail to explain the behav-
ior of emerging markets. We then allow for interest rates to respond to/co-vary with
productivity shocks. We find that emerging market business cycles appear to be driven
by large shocks to trend income combined with relatively small transitory shocks that
co-vary with the interest rate.∗[email protected], [email protected]. Prepared for the Tenth Annual Conference of the
Central Bank of Chile, “Current Account and External Financing,” to be held in Santiago, Chile, November
9 and 10, 2006. We thank Claudio Radatzz, Manuel Marfan and Kevin Cowan for very useful comments.
We also thank Deepa Dhume for excellent research assistance.
1
1 Introduction
Business cycles in Emerging Markets are characterized by high levels of volatility in income,
investment and net exports. Consumption is more volatile than income and net exports are
highly counter-cyclical. These facts are summarized in Aguiar and Gopinath (2007, hence-
forth AG (2007)). Further, the interest rates faced by these economies are highly volatile
and negatively correlated with income. These features of the interest rate process are sum-
marized in Neumeyer and Perri (2005). In this paper we adopt a standard stochastic small
open economy business cycle model and allow the economy to be driven by productivity
shocks that have permanent and transitory components as well as by shocks to the interest
rate process. We then estimate the role of the different processes in explaining the business
cycle behavior of emerging markets.
In AG (2007), we examined an economy driven exclusively by shocks to productivity.
Productivity shocks in this context may be viewed as manifestations of deeper frictions in
the economy such as changes in monetary, fiscal and trade policies. For instance, Restuccia
and Schmitz (2004) provide evidence of a 50% drop in productivity within five years in the
petroleum industry in Venezuela following its nationalization in 1975. Similarly, Schmitz
and Teixeira (2004) document almost a doubling of productivity in the Brazilian Iron-Ore
Industry following its privatization in 1991. It is such dramatic changes in productivity
following reforms and undoing of reforms that we view as characterizing emerging markets.
In addition, several emerging markets experience terms of trade shocks that display a high
degree of persistence. In this set-up we provided a methodology for identifying the role of
transitory versus trend shocks in explaining business cycles. The procedure relied on using
the intuition behind the permanent income hypothesis.
In AG (2007), we adopted the standard small open economy assumption and modelled
the interest rate as an exogenous international risk-free rate, which we held constant. In this
environment the economy always repays its debt and there is never any default. In Aguiar
and Gopinath (2006)we allow explicitly for default in an Eaton and Gersovitz set-up. We
2
specified an endowment economy driven by trend and stationary shocks. We show that
incorporating trend shocks is important in generating empirically plausible rates of default
as well as simultaneously matching key correlations between the interest rate, output, and
the current account.
In this paper, we extend AG (2007) to allow for a stochastic interest rate process. We
will consider three specifications. The first is the case of exogenous interest rate shocks that
are independent of the productivity shocks. The second is the case where in addition to
independent shocks the interest rate responds to transitory productivity shocks. The third
case is where the interest rate also responds to trend productivity shocks. We will assume
a reduced form specification for all these processes and provide intuition for the nature of
the process.
It is important to note that we estimate the interest rate process from the Euler Equa-
tions and do not use observed interest rates. This mirrors our treatment of productivity
shocks, where we do not use the Solow residual series to directly identify the underlying
productivity process. We do this for two reasons. Firstly, the observed rates are not risk free
rates given the probability of default. The promised rate observed in the data may therefore
not be the relevant real rate governing behavior.1 Secondly, agents may be constrained in
their access to financial markets. In that case, there is an implicit Lagrange multiplier that
governs the consumption/investment decision rather than the observed market rate. Our
estimation will pick up fluctuations in this multiplier. This approach is different from the
work of Neumeyer and Perri (2005) in that NP take the observed interest rate process and
feed this into the economy. This assumes that the euler equation with repayment is always
satisfied at the observed interest rates.
We show that the model with interest rate shocks that are orthogonal to productivity
shocks does poorly in matching the features of the data for emerging market countries.
Movements in the interest rate affect consumption and investment by setting the price for1See Aguiar and Gopinath (2006), Arellano (2006), Yue (2006) for explicit models of default.
3
intertemporal substitution. An increase in the interest rate reduces consumption relative
to the future as it increases the incentive to save. It also reduces investment in physical
capital as the return from the bond is higher. Since in this exercise interest rate shocks
are orthogonal to productivity shocks, the induced correlation between consumption and
income and investment and income is low, contrary to the data. The response of output,
on impact, to a rise in the interest rate will be small as productivity has not changed and
capital takes time to adjust. Moreover, when consumption and leisure are non-separable,
labor supply rises in response to a drop in consumption which generates an increase in
output, which is counterfactual given that high interest rate periods have been associated
with large declines in output. It is clearly the case that interest rate shocks that are not
associated with movements in productivity will perform poorly in matching the facts for
emerging markets. This point is similar in spirit to the work of Neumeyer and Perri (2005)
and Chari and Kehoe (2006).
We next allow the interest rate to respond to productivity shocks—both transitory and
trend shocks. The data suggest that a high level of productivity should be associated with
a lower interest rate. A positive shock to productivity raises consumption and the increase
is amplified by the contemporaneous decline in interest rates. This increases the relative
volatility of consumption for a given income process. Also, investment increases following
the rise in productivity and decline in interest rates. This implies that net-exports decrease,
inducing a negative correlation between net exports and income. The precise moments of
the stationary distribution will depend on the persistence in the income and interest rate
process. For reasons explained below, the model performs better when the interest rate
primarily responds to the transitory income shock.
Lastly, we use GMM and data from Mexico to estimate the parameters of a model that
allows for both exogenous interest rate shocks and productivity shocks and for the interest
rate shock to respond to the transitory income shock. In the benchmark case, where the
model allows only for productivity shocks, the random walk component of the Solow residual
is estimated to be 1.02. In AG (2007) we estimate that the random walk component for
4
Canada was far lower at 0.5. When we allow for the richer specification with interest rate
shocks, we estimate the random walk component to be essentially the same at 1.01. This
supports the conclusions in AG (2007) that emerging markets are subject to more volatile
trend shocks as compared to developed markets. We also find evidence of a small negative
covariance between productivity shocks and the implied interest rate.
It is important to emphasize that the differences in the Solow residual processes be-
tween developed and emerging markets may well be a manifestation of deeper frictions in
the economy. Chari, Kehoe, and McGrattan (forthcoming), for instance, show that many
frictions, including financial frictions, can be represented in reduced form as Solow residuals.
From the perspective of private agents in our economy, these shocks appear as exogenous
shifts in productivity. Our analysis provides support for models with frictions that are re-
flected in the persistence of Solow residuals, rather than frictions that distort the response
of investment and consumption to underlying productivity. For instance, Guajardo (2007),
describes a model with financial frictions and finds that it fits the data best when there are
exogenous labor financing wedges which affect hiring decisions and are procyclical. That is,
financing working capital requirements is easier in booms as compared to recessions. These
financing wedges behave like productivity shocks. Our analysis shows that interest rate
shocks that only affect the Euler Equation add little to matching the facts in the data for
emerging markets. Clearly, one could argue that there are interest rate movements that
interact with underlying financial frictions to generate shocks that look like productivity
shocks. Our analysis is completely consistent with such a model.
We also present here evidence that Chile has features similar to other emerging markets
documented in AG (2007).2 The correlation between HP-filtered net exports as a ratio
of GDP and HP-filtered log of GDP is -0.82 for Chile. There does not exist a quarterly
series on private consumption before 1996. For the 10 years from 1996-2006 the volatility
of HP-filtered log GDP is 1.63 compared to a volatility of 1.89 for HP-filtered log of private
consumption. This is similar to other emerging markets, which on average experience2We thank David Rappoport for providing us with this data
5
consumption volatility that exceeds the volatility of income and net exports that are highly
counter-cyclical. In Section 2 we describe the stochastic growth model. In Section 3 we
describe the identification strategy and provide intuition by presenting impulse responses
to various shocks. Lastly, in Section 4 we describe the results from a GMM estimation of
the model.
2 Stochastic Growth Model
The model here is reproduced from AG (2007) and augmented to include a stochastic
interest rate process. Technology is characterized by a Cobb-Douglas production function
that uses capital, Kt, and labor, Lt, as inputs
Yt = eztK1−αt (ΓtLt)α, (1)
where α ∈ (0, 1) represents labor’s share of output. The parameters zt and Γt represent pro-
ductivity processes. The two productivity processes are characterized by different stochastic
properties. Specifically, zt follows an AR(1) process
zt = ρzzt−1 + εzt (2)
with |ρz| < 1, and εzt represents iid draws from a normal distribution with zero mean and
standard deviation σz.
The parameter Γt represents the cumulative product of “growth” shocks. In particular,
Γt = egtΓt−1 =t∏
s=0
egs
gt = (1− ρg)µg + ρggt−1 + εgt ,
where |ρg| < 1 and εgt represents iid draws from a normal distribution with zero mean and
standard deviation σg. The term µg represents productivity’s long-run mean growth rate.
We loosely refer to the realizations of g as “growth shocks” as they constitute the stochastic
6
trend of productivity. We use separate notation for shocks to the level of productivity (zt)
and the growth of productivity (gt) to simplify exposition and calibration.
Given that a realization of g permanently influences Γ, output is nonstationary with a
stochastic trend. For any variable x, we introduce a hat to denote its detrended counterpart:
xt ≡ xt
Γt−1.
Note that we normalize by trend productivity through period t − 1. This insures that if
xt is in the agent’s information set as of time t − 1, so is xt. The solution to the model is
invariant to the choice of normalization.
Period utility is Cobb-Douglas,
ut =
(Cγ
t (1− Lt)1−γ
)1−σ
1− σ(3)
where 0 < γ < 1. A well-behaved steady state of the deterministic linearized model requires
β(1 + r∗) = µ1−γ(1−σ)g .
The equilibrium is characterized by maximizing the present discounted value of utility
subject to the production function (1) and the per-period resource constraint:
Ct + Kt+1 = Yt + (1− δ)Kt − φ
2
(Kt+1
Kt− eµg
)2
Kt −Bt + qtBt+1. (4)
Capital depreciates at the rate δ and changes to the capital stock entail a quadratic adjust-
ment cost φ2
(Kt+1
Kt− eµg
)2Kt. We assume international financial transactions are restricted
to one-period, risk-free bonds. The level of debt due in period t is denoted Bt and qt is the
time t price of debt due in period t + 1.
Fluctuations in the price of debt, qt, will be our focus. We assume that the interest
rate is potentially driven by an exogenous process rt as well as the domestic TFP shocks.
Specifically, the price of debt q is given by the expression below.
7
1qt
= 1 + r∗ + e{rt+azzt+ag(gt−µg)} + ψ
[e
Bt+1Γt
−b − 1]
, (5)
where
rt = ρrrt−1 + εrt . (6)
The world interest rate is held constant at r∗. The country-specific shock to the interest rate
is given by εrt , which is orthogonal to z and g. The induced process rt has an autocorrelation
coefficient ρr and a long run mean of zero. The parameters az and ag capture the sensitivity
of the interest rate to the the transitory productivity shock and the trend productivity shock,
respectively. We should note that the correlation of the interest rate and productivity does
not imply a direction of causation between the two. See Aguiar and Gopinath (2006) for a
model in which exogenous domestic productivity shocks drive an endogenous interest rate,
while Neumeyer and Perri (2005) present a model in which exogenous (foreign) interest rate
shocks drive domestic TFP. b represents the steady-state level of debt, and ψ > 0 governs
the elasticity of the interest rate to changes in indebtedness. This sensitivity to the level
of outstanding debt, takes the form used in Schmitt-Grohe and Uribe (2003).3 In choosing
the optimal amount of debt, the representative agent does not internalize the fact that she
faces an upward-sloping supply of loans.
In normalized form, the representative agent’s problem can be stated recursively:
V (K, B, z, g, r) = max{C,L,K′,B′}
(Cγ (1− L)1−γ
)1−σ
1− σ+ βegγ(1−σ)EV (K ′, B′, z′, g′, r′)
(7)
3This adjustment is typically motivated by the need to make assets in the linearized model stationary. An
alternative is to recognize that we are linearly approximating a non-linear economy for which a stationary
distribution exists (for example, due to borrowing constraints and a world equilibrium interest that is lower
than the discount rate, as in Aiyagari 1994). Quantitatively, since the elasticity of interest rate to changes in
indebtedness is set close to 0 (0.001 to be exact), there is a negligible difference between the two approaches
in terms of the HP-filtered or first-differenced moments of the model.
8
s.t. C + egK ′ = Y + (1− δ)K − φ
2
(eg K ′
K− eµg
)2
K − B + egqB′. (8)
The evolution of the capital stock is given by,
egK ′ = (1− δ) K + X − φ
2
(K ′
Keg − eµg
)2
K. (9)
Given an initial capital stock, K0, and debt level, B0, the behavior of the economy is
characterized by the first-order conditions of the problem (7), the technology (1) and budget
(8) constraints, and the transversality conditions.
We solve the normalized model numerically by log-linearizing the first-order conditions
and resource constraints around the deterministic steady state. Given a solution to the
normalized equations, we can recover the path of the non-normalized equilibrium by mul-
tiplying through by Γt−1. We also compute the theoretical moments of the model from the
coefficients of the linearized solution.
3 Identification
The primary goal of this paper is to assess the relative importance of interest rate shocks,
transitory productivity shocks and permanent shocks to productivity in explaining the
behavior of emerging markets. In AG (2007) we described the methodology of exploiting
decisions by informed, optimizing agents to identify the underlying shock process. In this
paper, we extend that methodology to accommodate a richer process for the interest rate.
The methodology we employ selects parameters of the model to match key moments
of the data. Below, we discuss which moments are particularly useful in identifying the
parameters of interest. However, we note from the start that we do not use market interest
rates on sovereign debt. The reason is that those interest rates represent the price of a
defaultable bond. This is a different asset than that modelled above. To see this, consider
9
the Euler Equation for bonds from the above model:
β
qE
uc′
uc= 1. (10)
Note that while consumption is stochastic, the interest rate paid (conditional on information
at the time of borrowing) is deterministic. In a model with defualtable debt, the consumer
pays the interest rate conditional on no-default, and pays zero (or some fraction) if default
occurs. Therefore, the observed market interest rate cannot be used directly in a simple
Euler Equation, but must be combined with a full specification of in which states default
occurs and what payments are made conditional on default.
Our interest rate process q can be viewed as a wedge in the Euler Equations for con-
sumption and investment. Our estimation will then back out the parameters governing
the stochastic process of this wedge. In this sense, it is similar to the exercise of Chari,
Kehoe, and McGrattan (forthcoming). It also captures unobserved frictions (to a linear ap-
proximation) such as additional borrowing costs or constraints beyond the market interest
rate.
3.1 Interest Rates Shocks Orthogonal to Productivity Shocks
We begin with an exploration of uncorrelated interest rate shocks. That is, shocks to
the interest rate that are orthogonal to total factor productivity. Changes to the inter-
est rate induce changes to consumption and investment for a given path of income due to
inter-temporal substitution. This will raise the relative volatility of consumption and in-
vestment. Therefore, such shocks have the potential to explain the relatively high volatility
of consumption in emerging markets.
However, by introducing shocks that move consumption and investment independently of
income reduces the covariance of consumption and investment with income. This generates
counterfactual implications for the cyclicality of net exports.
Figure 1 plots the impulse response of consumption, investment, net exports, and income
10
to a one percent shock to εr. We set ρr = 0.9. As expected, an increase in the interest rate
leads to a drop in consumption, with an initial decline of roughly 3 percent. Investment
declines in an even more dramatic fashion. Output remains steady, declining slightly over
the path due to the lagged declines in investment. This leads to a jump in net exports.
To see how orthogonal interest rate shocks affect key moments of the simulated model,
consider a model in which we set az = ag = 0, but set σr ≡ stdev(εr) > 0. To be precise, we
consider models with various σr ranging from zero to one percent. For each environment,
we compute key moments of the simulated economy and plot them in Figure 2. We fix all
other parameters. We also set γ = 1 so that labor supply is fixed. All moments refer to HP-
filtered variables. In Panel A of Figure 2 we see how the relative variance of consumption,
investment, and net exports increases as we increase σr. This corresponds to the above
intuition. In Panel B, we see that net exports become more pro-cyclical as we increase σr.
This takes us further from the data. Correspondingly, consumption, investment become less
correlated with income. This is because a positive interest rate shock lowers consumption
and investment. As TFP has not changed, this reduces the correlation with income. In fact,
when consumption and leisure are nonseparable, the decreased consumption is associated
with higher labor and therefore higher income, inducing a negative correlation between
consumption and income. In this set-up, a crisis which is associated with a large increase
in interest rates, will reduce consumption but raise output, which is completely counter-
factual.
It is clearly the case that exogenous interest rate shocks will do poorly in explaining
the behave of emerging markets. It will be hard to generate the large counter-cyclicality in
the current accounts and the much larger responsiveness of consumption relative to income.
This argument is in line with the results in Neumeyer and Perri (2005) and Chari and Kehoe
(2006). A model where the interest rate process does not show up as affecting productivity
will have little hope of matching moments of the business cycle.
11
3.2 Interest Rate that Co-vary with Productivity Shocks
From the previous section it is clear that we need to interact the interest rate shock with
the productivity shock. As we have two productivity processes, there are two dimensions
along which we can link the interest rate and productivity. We begin by setting ag = 0 and
considering the link between transitory productivity shocks and the interest rate. We then
set az equal to 0 and assume the interest rate responds only to the permanent shock g.
To gain some intuition, in Figure 3 we plot the impulse response of consumption and
income to a shock to εz when az = 0 and when az = −0.1. This latter case generates a fall
in the interest rate when productivity increases. This can be an implication of the Eaton
Gersovitz style models of default in which default occurs during low income realizations (see
Aguiar and Gopinath (2006) and Arellano (2004)). With persistent shocks, a high shock
today implies on average high shocks tomorrow and a correspondingly low probability of
default. This generates a negative relationship between productivity and the interest rate.
For the benchmark case of az = 0, we see the standard consumption smoothing result —
consumption increases, but income increases much more. The case of az < 0 combines the
income response with a substitution response that favors initial consumption. This generates
a larger initial jump in consumption and a subsequent declining profile of consumption.
Given the transitory nature of the shock, the net effect is that consumption tracks the
shape of the income impulse response. Not depicted is the response of investment, which
has a similar intuition as consumption.
The impulse responses indicate that allowing the interest rate and productivity to co-
move overcome some of the limitations of transitory productivity. Namely, consumption
and investment respond more to income and in a way that makes net exports negatively
associated with income. To see how this extension affects business cycle moments, we plot
the key moments as a function of az in Figure 4. As az becomes increasingly negative,
this raises the volatility of consumption relative to income. A positive productivity shock
lowers interest rates, generating an increase in consumption above and beyond the income
12
effect. Unlike the orthogonal interest rate process of Figure 2, the additional consumption
volatility increases the correlation of consumption and income. This effect is driven by the
fact that the interest rate moves one-for-one with productivity. A similar story holds for
investment. These effects make net export countercyclical, a key feature of the data for
emerging markets.
As noted above, an alternative approach is to allow the interest rate to respond to
permanent productivity shocks, i.e. ag < 0. In Figure 5 we plot the impulse response to
a shock to εg in the benchmark case and in the case when ag = −1. Given that g has a
permanent effect on income, we see that consumption responds strongly to the initial shock
in the benchmark case, exceeding the initial response of income. Allowing the interest rate
to respond as well heightens the initial response of consumption. However, subsequently,
the interest falls back quickly to its initial level as g is nearly iid. This generates a sharp
fall in consumption and then a levelling out. However, income jumps and then continues
to rise in response to a growth shock. Therefore, allowing ag < 0 lowers the correlation of
consumption with income, taking us further from the data.
This effect can be seen clearly in Figure 6. As we increase ag (in absolute value),
while the variance of consumption and investment increase, the correlations with income
at business cycle frequencies fall. This reduces the cyclicality of net exports, drawing us
further from the data.
The poor performance of the model with ag < 0 is due to the fact that growth rates have
little persistence, generating interest rates with little persistence. An alternative would be
that interest rates are a function of the level of the stochastic trend Γ. However, this would
imply a nonstationary interest rate.
13
3.3 Productivity Shocks Alone
AG (2007) considered a model in which ag = az = 0. We briefly summarize the intuition
behind the identification of the relative variance σg/σz. In response to a transitory shock
to productivity, agents increase consumption, but by less than the increase in income since
they expect income to be lower in the future and by saving they can smooth consumption.
On the other hand, if the economy is hit by a growth shock which implies permanently
higher income and depending on the persistence of the growth shock an upward sloping
profile of income the agents will increase consumption by at least as much as the increase
in income. Therefore consumption is more volatile relative to income in the world with
permanent shocks relative to transitory shocks. This difference in the response of σc is
observed in figure 7.
Therefore, by observing the behavior of consumption we can infer the relative impor-
tance of trend versus transitory shocks. Similarly, it follows that given the response of
consumption and we should expect net exports to be far more countercyclical for the econ-
omy with trend shocks and the moment on net exports can be used to identify the underlying
productivity shock.
3.4 Identification Strategy
Given the above results, we restrict σr = ag = 0. That is, we consider a model in which
the interest rate co-varies with transitory productivity shocks and allow for both transitory
and trend shocks to productivity. The patterns depicted in figures 4 and 7 indicate how we
can identify the key parameters. Increases in the magnitude of az and σg/σz have a similar
impact on the cyclicality of the current account. However, while both raise the relative
volatility of consumption, net exports, and investment, the relationships differ. Figure 4
indicate that az has an almost linear effect on the relative variances, while figure 7 indicates
that the impact of σg/σz eventually dies out. In particular, for large enough az, the relative
14
volatility of net exports exceeds that of consumption. This reflects the differential sensitivity
of investment and consumption to interest rate shocks. Therefore, the empirical moments
of σ(c) and σ(nx) combined with the empirical covariance of net exports with output pin
down the relative magnitudes of az and σg/σz. Given the relative variance of trend and
transitory shocks, the level of income volatility then identifies the level of σz and σg.
4 Estimates
Following the above identification strategy, in this section we estimate σg, σz, and az by
matching the following (HP-filtered) moments of the data: the standard deviations of in-
come, consumption, and net exports, as well as the covariance of net exports with income.
We use data from Mexico as a representative emerging market and Canada as a repre-
sentative developed open economy. We fix other parameters at the values listed in Table
5
For each set of estimates, we report the relative importance of the random walk com-
ponent of productivity. Beveridge-Nelson (1981) showed that any I(1) series can be decom-
posed into a random walk component (denoted τ) and a stationary component. A natural
measure of the importance of the random walk component is the ratio of the variance of
the growth rate of the trend component to the growth rate of total TFP.
σ2∆τ
σ2∆TFP
=α2σ2
g
(1− ρg)2σ2∆TFP
=
α2σ2g
(1−ρg)2
[ 2σ2z
1+ρz+ α2σ2
g
1−ρ2g]
(11)
We report the estimates for σg, σz and az in Table 5. In the columns denoted “bench-
mark” we restrict az = 0. This corresponds to the benchmark model of AG (2007). The
other columns estimate az. The first two columns consider a model in which labor is sup-
plied exogenously. This corresponds to setting the Cobb-Douglas preference parameter on
consumption (γ) to one, so that leisure does not enter utility. The next two columns allow
labor supply to vary endogenously, setting γ = 0.36. The final two columns estimate the
15
model using Canadian data.
For the benchmark model using Mexican data (column 1), we see that σg is larger than
σz and that the relative contribution of the random walk component to TFP is 1.02. This is
similar to the results reported in AG (2007). In the second column of Table 2, we estimate
az along with σz and σg. We find that az < 0, although we cannot reject az = 0 at standard
significance levels. Even allowing for interest rate shocks, we estimate a relatively large σg,
with the contribution of the random walk component estimated to be 1.01.
Allowing labor supply to vary endogenously does not overturn this pattern. In both
specifications, the random walk component of productivity is estimated to be roughly 1.0.
The coefficient az is estimated to be small.
The case of Canada indicates a relatively small random walk component. In both
specifications, we estimate the relative random walk component to be 0.4. The coefficient
az is also estimated to be small and not significantly different from zero.
In Table 5 we report the implied business cycle moments from the estimated models
along with the corresponding empirical moments from Mexico and Canada. The implied
moments for Mexico correspond to the first two columns of Table 5. In respect to Mexico,
we see that both models perform well in matching key features of the data. The empirical
relative volatility of consumption is 1.3, while the models with and without interest rate
shocks both generate relative variances of 1.1. The cyclicality of net exports is -0.8 in the
data and is -0.7 and -0.6 in the models without and with interest rate shocks, respectively.
In general, allowing for interest rate shocks does not markedly improve the fit of the model.
A similar story holds for Canada, as seen in the final columns of Table 5.
In the specification with interest rate shocks, we see that interest rates are countercyclical
in Mexico and procyclical in Canada. However, the variance of the implied interest rates
is negligible. This reflects that while consumption is volatile in emerging markets, it is not
driven by inter-temporal substitution, but rather by income shocks.
16
5 Discussion and Conclusion
Emerging Markets are characterized by large volatility in their income and consumption
and large countercyclicality in net exports as compared to developed small open economies.
They also face a volatile interest rate process that is negatively correlated with the level of
GDP in these economies. There is a large amount of literature that attempts to explain
these features of the data and infer the importance of productivity and interest rate shocks
in explaining the patterns observed in the data. In this paper we perform a similar exercise
by extending the framework in AG(2007) which allowed only for productivity shocks to one
that allows for a richer specification of interest rate shocks and for the interaction between
productivity and interest rate shocks.
One finding, which supports other evidence in the literature, is that interest rate shocks
that do not effect productivity cannot be the main explanation for the business cycles
of emerging markets. These markets are characterized by large movements in output at
business cycle frequencies that are associated with large movements in the Solow resid-
ual. Accordingly, interest rate shocks alone will do little to explain these large movements
in output. It is important to uncover channels through which interest rate shocks affect
productivity.
If the interest rate is negatively correlated with the productivity shock, then interest
rates can indeed play an important role. It can explain, at least qualitatively, a consump-
tion process that is more volatile than income and counter-cyclical net exports. When we
estimate the model to allow for the interaction between interest rates and productivity we
find a small negative correlation between productivity and interest rates. We also find that,
even in this framework, we obtain a large role for trend shocks which supports the main
result in AG (2007), namely, that an important characteristic of emerging markets is that
shocks to trend productivity play a predominant role in explaining movements at business
cycle frequencies, unlike developed markets.
17
In this paper we have taken a reduced form approach to modelling both the interest
rate process and productivity shocks. Future work should examine the structural features of
emerging markets that give rise to the particular form of these processes. In AG (2006) we
explore a model with Eaton-Gersovitz style endogenous default. While this approach does
generate default in equilibrium and can generate a countercyclical interest rate process, it
fails to generate sufficient volatility in the market interest rate process. Further research is
required to understand the source of the volatility in the interest rate process.
18
References
[1] Aguiar, Mark and Gita Gopinath (2006). “Defaultable Debt, Interest Rates and the
Current Account.” Journal of International Economics, 69(1): 64-83.
[2] Aguiar, Mark and Gita Gopinath (2007). “Emerging Market Business Cycles: The
Cycle is the Trend,” Journal of Political Economy, Vol 115,1, 69–102.
[3] Aiyagari, S. Rao (1994), “Uninsured Idiosyncratic Risk and Aggregate Saving,” The
Quarterly Journal of Economics, Vol 109, 3, 659-684.
[4] Arellano, Cristina and Enrique Mendoza (2002). “Credit Frictions and ‘Sudden Stops’
in Small Open Economies: An Equilibrium Business Cycle Framework for Emerging
Market Crises.” NBER Working Paper No. w8880.
[5] Backus, David, Patrick J. Kehoe and Finn. E. Kydland (1995). “International Business
Cycles: Theory and Evidence.” Frontiers of Business Cycle Research. ed. Thomas F.
Cooley. Princeton University Press. Princeton, New Jersey.
[6] Bergoeing, Raphael, Timothy Kehoe, Patrick Kehoe and Raimundo Soto (2002). “A
Decade Lost and Found: Mexico and Chile in the 1980s.” Review of Economic Dynam-
ics, 5 , 166-205.
[7] Bernanke, Ben and Refet Gurkaynak (2002). “Is Growth Exogenous? Taking Mankiw,
Romer and Weil Seriously.” NBER Macroeconomics Annual 16:11-57.
[8] Beveridge, Stephen, and Charles R. Nelson (1981). “A New Approach to Decomposition
of Economic Time Series into Permanent and Transitory components with Particular
attention to the Measurement of the Business Cycle.” Journal of Monetary Economics
7(2):151-74.
[9] Blundell, Richard and Ian Preston (1998), “Consumption Inequality and Income Un-
certainty,” Quarterly Journal of Economics, 113(2): 603-640.
19
[10] Calvo, Guillermo A. (1986), “Temporary Stabilization: Predetermined Exchange
Rates,” Journal of Political Economy, 94(6): 1319-1329.
[11] Calvo, Guillermo A. and Carmen Reinhart (2000). “When Capital Inflows come to
a Sudden Stop: Consequences and Policy Options.” in Peter Kenen and Alexandre
Swoboda. “Key Issues in Reform of the International Monetary and Financial System.”
Washington DC: International Monetary Fund: 175-201.
[12] Campbell, John and Angus Deaton (1989). “Why is Consumption so Smooth?” Review
of Economic Studies. 56(3) July:357-373.
[13] Chari, V.V., Patrick Kehoe and Ellen McGrattan (forthcoming), “Business Cycle Ac-
counting, ” Econometrica.
[14] Cochrane, John (1988). “How Big is the Random Walk in GNP?” Journal of Political
Economy 96(5) October:893-920.
[15] Cochrane, John (1994). “Permanent and Transitory Components of GNP and Stock
Prices.” Quarterly Journal of Economics 109(1) February:241-265.
[16] Correia, Isabel, Joao C. Neves, Sergio Rebelo (1995). “Business Cycles in Small Open
Economies.” European Economic Review 39(6) June:1089-1113.
[17] Dornbusch, Rudiger and Sebastian Edwards (1992). “Macroeconomics of Populism in
Latin America.” NBER Conference Volume.
[18] Greenwood, Jeremy, Zvi Hercowitz and Gregory W. Hoffman (1988). “Investment,
Capacity Utilization, and Real Business Cycle.” American Economic Review 78(3)
June:402-17.
[19] Hamilton, James (1994), Time Series Analysis, Princeton University Press, Princeton,
NJ.
[20] Hansen, Lars (1982). “Large Sample Properties of Generalized Method of Moments
Estimators.” Econometrica 50(4):1029-1054.
20
[21] Mendoza, Enrique (1991). Real Business Cycles in a Small Open Economy. American
Economic Review 81(4) September:797-818.
[22] Neumeyer, Pablo A. and Fabrizio Perri (2005). “Business Cycles in Emerging Markets:
The Role of Interest Rates.” Journal of Monetary Economics 52(2): 345-380.
[23] Restuccia, Diego and Schmitz, James (2004) ”Nationalization’s Impact on Output and
Productivity: The case of Venezuelan Minerals”, working paper.
[24] Schmitt-Grohe, Stephanie and Martin Uribe (2003). “Closing Small Open Economy
Models.” Journal of International Economics 61(1) October:163-185.
[25] Schmitz, James and Teixeira, Arilton (2004), “Privatization’s Impact on Private Pro-
ductivity: The case of Brazilian Iron Ore”, working paper.
[26] Standard and Poor’s, (2000). Emerging Stock Markets Factbook 2000 McGraw Hill
Company. New York, NY.
[27] Stock, James H. and Mark W. Watson (2003). “ Understanding Changes in Interna-
tional Business Cycle Dynamics.” NBER Working Paper no. w9859.
21
Table 1: Benchmark Parameter Values
Time Preference Rate β 0.98
Coefficient of Relative Risk Aversion σ 2
Cobb-Douglas Utility Parameter γ 1, 0.36
Steady State Debt to GDP b 10%
Coeff. on Interest Rate Premium Ψ 0.001
Labor Exponent (Production) α 0.68
Depreciation Rate δ 0.05
Capital Adjustment Cost φ 1.5
Persistence in z process ρz 0.95
Persistence in g process ρg 0.01
22
Table 2: Estimates
Mexico Canada
Exogenous Labor Endogenous Labor Endogenous Labor
Benchmark With az Benchmark With az Benchmark With az
σz 0.13 0.16 0.13 0.24 0.72 0.69
(2.42) (0.79) (0.66) (1.06) (0.09) (0.16)
σg 2.78 2.70 2.69 2.68 0.84 0.89
(0.44) (0.33) (0.00) (0.31) (0.15) (0.09)
az -0.40 -0.01 0.01
(1.85) (0.55) (0.02)
Random Walk 1.02 1.01 1.01 1.00 0.39 0.44
Component (0.18) (0.08) (0.05) (0.15) (0.07) (0.13)Notes: Estimates obtained from matching empirical moments of Mexico and Canada for respective columns. Moments
used were standard deviation of HP-filtered log income, log consumption, and net exports/GDP as well as the
covariance of HP-filtered net exports/GDP and log income. Exogenous Labor model sets γ = 1. Endogenous Labor
model sets γ = 0.36.
23
Table 3: Implied Moments
Mexico Canada
Data Model I Model II Data Model I Model II
σ(y) 2.40 2.69 2.63 1.55 1.56 1.55
σ(c)/σ(y) 1.26 1.09 1.10 0.74 0.71 0.72
σ(nx)/σ(y) 0.90 0.74 0.84 0.57 0.59 0.60
σ(i)/σ(y) 4.15 3.52 3.81 2.67 3.23 3.13
σ(r) NA 0.08 NA 0.01
ρ(yt, yt−1) 0.82 0.78 0.78 0.90 0.79 0.79
ρ(c, y) 0.92 0.98 0.98 0.87 0.87 0.85
ρ(nx, y) -0.75 -0.68 -0.61 -0.12 -0.17 -0.13
ρ(i, y) 0.91 0.86 0.82 0.74 0.85 0.84
ρ(r, y) NA -0.01 NA 0.90Notes: Empirical moments and implied moments from alternative models. Model I and Model II
for Mexico correspond respectively to the first two columns of estimates (exogenous labor supply
model) of Table 2. Model I and Model II of Canada correspond to the respective columns of
estimates for Canada from Table 2.
24
Figure 1: Impulse Response to Interest Rate Shock
-20
-15
-10
-5
0
5
10
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
cyinx
Note: Impulse response to a 1 percent shock to εr
Figure 2: Business Cycle Moments and σr
Panel A: Standard Deviation of Investment, Consumption, and Net Exports
0.0
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
0.00% 0.10% 0.20% 0.30% 0.40% 0.50% 0.60% 0.70% 0.80% 0.90% 1.00%
stdev of log interest rate
stde
v (r
el to
inco
me) investment consumption
net exports
Note: Standard Deviation of (HP-filtered, log) consumption, investment, and net exports relative to income as a function of σr.
Panel B: Cyclicality of Investment, Consumption, and Net Exports
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
-0.03 0.04 0.07 0.10 0.11 0.12 0.12 0.13 0.13 0.13 0.13
stdev of log interest rate
corr
with
inco
me
investment consumption net exports
Note: Correlation of (HP-filtered, log) consumption, investment, and net exports with income as a function of σr.
Figure 3: Impulse Response to z Shock
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
consumption_benchmarkincome_benchmarkconsumption_azincome_az
Note: Impulse response to a 1 percent shock to εz. Benchmark model sets az=0. “az” model sets az=-0.1.
Figure 4: Business Cycle Moments and az Panel A: Standard Deviation of Investment, Consumption, and Net Exports
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
0.00 -0.01 -0.02 -0.03 -0.04 -0.05 -0.06 -0.07 -0.08 -0.09 -0.10
a_z
stde
v (r
el to
inco
me)
investmentconsumptionnet exports
Note: Standard Deviation of (HP-filtered, log) consumption, investment, and net exports relative to income as a function of az.
Panel B: Cyclicality of Investment, Consumption, and Net Exports
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0.00 -0.01 -0.02 -0.03 -0.04 -0.05 -0.06 -0.07 -0.08 -0.09 -0.10
a_z
corr
with
inco
me
investment consumption
net exports
Note: Correlation of (HP-filtered, log) consumption, investment, and net exports with income as a function of az.
Figure 5: Impulse Response to g Shock
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
consumption_benchmarkincome_benchmarkconsumption_agincome_ag
Note: Impulse response to a 1 percent shock to εg. Benchmark model sets ag=0. “ag” model sets ag=-1.
Figure 6: Business Cycle Moments and ag
Panel A: Standard Deviation of Investment, Consumption, and Net Exports
0.0
2.0
4.0
6.0
8.0
10.0
12.0
0.00 -0.10 -0.20 -0.30 -0.40 -0.50 -0.60 -0.70 -0.80 -0.90 -1.00
a_g
stde
v (r
el to
inco
me)
investment consumption
net exports
Note: Standard Deviation of (HP-filtered, log) consumption, investment, and net exports relative to income as a function of ag.
Panel B: Cyclicality of Investment, Consumption, and Net Exports
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0.00 -0.10 -0.20 -0.30 -0.40 -0.50 -0.60 -0.70 -0.80 -0.90 -1.00
a_g
corr
with
inco
me
investmentconsumptionnet exports
Note: Correlation of (HP-filtered, log) consumption, investment, and net exports with income as a function of ag.
Figure 7: Business Cycle Moments and σg/σz
Panel A: Standard Deviation of Investment, Consumption, and Net Exports
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
sigma(g)/sigma(z)
stde
v (r
el to
inco
me)
investmentconsumptionnet exports
Note: Standard Deviation of (HP-filtered, log) consumption, investment, and net exports relative to income as a function of σg/σz.
Panel B: Cyclicality of Investment, Consumption, and Net Exports
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
0.00 -0.10 -0.20 -0.30 -0.40 -0.50 -0.60 -0.70 -0.80 -0.90 -1.00
sigma(g)/sigma(z)
corr
with
inco
me
investmentconsumptionnet exports
Note: Correlation of (HP-filtered, log) consumption, investment, and net exports with income as a function of σg/σz.